Background:
Adiponectin (APN) is inversely related to incident cardiovascular disease (CVD) in healthy middle-aged cohorts, but the opposite has been observed among older populations or those with prevalent CVD, where higher APN imparts greater risk of CVD and death. Emerging data suggest, however, that the association of total APN with mortality in elders may be U-shaped.
Methods:
We tested the hypotheses that both total and high-molecular-weight (HMW) APN (r=0.94) manifest different relations with mortality in subgroups of older adults defined by the presence or absence of prior CVD or heart failure (HF)/atrial fibrillation (AF). Specifically, we hypothesized that total and HMW APN would show similar U-shaped associations with all-cause and CVD death in subjects without prevalent CVD or HF/AF (Group [Gp] 1; n= 3272), but would exhibit positive monotonic associations with these outcomes in subgroups with prevalent CVD but no HF/AF (Gp 2; n=1030), and with prevalent HF/AF (Gp 3; n=383). We addressed these questions in CHS, a population-based US cohort aged 65 and older, of whom 4715 had available samples since 1992–93. Associations were examined with general additive model plots, and modeled with linear splines.
Results:
During 16 years of follow-up, 1947 all-cause and 634 CVD deaths occurred in Gp 1, 802 and 375 in Gp 2, and 337 and 180 in Gp 3. There was evidence of effect modification by subgroup status for both outcomes (p≤0.034), with total and HMW APN showing significant departures from linearity in their relations with all-cause and CVD mortality in Gp 1 (p≤0.043), but not Gps 2 or 3. The association between total APN and all-cause mortality was U-shaped, such that after adjustment for potential confounders, increasing levels up to 12.4 mg/L (median) were associated with a lower risk of death (HR 0.81 per SD [0.65–0.95]), but above this cutpoint, higher levels imparted a higher risk (HR 1.19 per SD [1.12–1.27]). Further adjustment for putative mediators (glucose, lipids, inflammation) abolished the association in the lower range, but left that in the upper range unaffected. The relationship was largely similar for HMW adiponectin. No significant association between total or HMW APN with mortality was apparent in Gp 2. In Gp 3, both total and HMW APN showed positive adjusted associations with mortality across their distributions, which were magnified after inclusion of putative mediators (HRs 1.31 [1.15–1.50] and 1.36 [1.20–1.55], respectively). Results were comparable for CVD mortality in all Gps.
Conclusions:
These findings show that total and HMW APN bear similar associations with all-cause and CVD mortality in older adults, and that these differ according to prevalent CVD or HF/AF status. These observations provide a potential explanation for the APN paradox, underscoring the need to better characterize the underpinnings of the hormone’s beneficial and harmful associations.
Inverse correlation between serum high‐molecular‐weight adiponectin and proinsulin level in a Japanese population: The Dynamics of Lifestyle and Neighborhood Community on Health Study
